Intel has finally confirmed the existence of Yamhill. To quote:
Intel spokesman Robert Manetta says the Santa Clara, Calif.-based chipmaker has a working prototype of a 64-bit x86 design that it could bring to market “when customers request it.”
This is Intelese for “we’re going to end up doing this.”
Why Did Intel Change Its Mind?
There’s several probable reasons that made Intel change its mind:
The Ramp Is Broken
As we said last week, Intel is having a real problem with its circuitry which seems to be seriously affecting its ability to ramp up frequencies like they’ve been accustomed to doing.
It’s not clear what the exact nature of the problem is (the complete answer is probably rather complex), but here, the what is more important than the why.
x86-64 gives you on average roughly 20% more performance. That’s a very erratic 20%, by the way, a few apps could show a 50%+ improvement, but many if not most typical apps will hardly show any improvement (as we explained a while back.
For the average computer user, it’s not a must-have, and for that average user, they’re much better off with a 32-bit running a lot faster than with x86-64.
However, if you can’t ramp frequency as easily or by as much as you could previously, it’s not so easy to turn your nose up at the benefits of x86-64. You can laugh at a 2.8 or 3 GHz x86-64 when you can easily make a 5-6GHz PIV. You can’t when you’re struggling to make a 4GHz PIV, because that’s going to lose against the x86-64.
It is this reason that turned the hard heads at Intel, simply because it legitimizes the BS PR reasons for x86-64.
Never Underestimate The Intelligence of the Average Computer Buyer
Over the course of time, I’ve gotten more than a few emails essentially saying, “the masses will buy x86-64 because 64 is more than 32.”
Well, they’re usually not quite that dumb. Or perhaps these folks have forgotten the counter-dumbness called the MHz Myth.
Yes, 64 is more than 32, but 5GHz is more than 3GHz. Oh my, what’s a mentally challenged computer buyer to do?
What that mentally challenged computer buyer will do is probably look at the media, and the media will look at who wins the benchmarking contests to see which dumbness wins out.
If Intel wins the contests (i.e., can ramp up normally), then the media will say that MHz is what matters, and 64 is BS. If Intel can’t ramp, and loses to Hammers, the generally-ignorant media will say that proves 64 is better than 32, and forget the MHz stuff.
So you see, 64-bitness all by itself isn’t a winner, but 64-bitness and winning races gives it traction.
Neither of the media’s conclusion would be true, but it will become the “truth” in the minds of millions. Intel most definitely doesn’t want that to happen. The MHz Myth has served it long and well.
x86-64 and IA-64: The Real Race
Before one kneejerks a denunciation of Itanium and IA-64, one ought to consider why Intel went off on this venture in the first place. There are two major reasons.
First, x86 is an OLD architecture. It was initially designed for the 8086 twenty-five years ago, and was meant for 16/8-bit processors. To keep compatibility with that relic, you have to do lots of less-than-optimal circuit designs.
Intel saw that they couldn’t do this indefinitely, and launched what later became the Itanium to get a fresh start, and toss out a lot of old to make room for new and better.
In short, Intel said, “The horse is getting old. We need a new horse.”
AMD couldn’t afford to buy a new horse, and Intel wouldn’t let them ride theirs, so they essentially said, “Let’s buy a new whip and get one more ride out of the old nag.”
Second, in the early eighties, Intel signed licenses with AMD that let them make x86 processors. That’s why AMD can make them. There are no such agreements about IA-64. AMD would have no clear legal right to make CPUs based on that new architecture, so if the world went to IA-64, AMD would be shut out of it unless they could sue their way into it (which could well fail, or at the least take a lot of time).
As you can see, there are really two questions here. The first is, “Do we have to replace the old horse?” The second is, “When do we have to replace the old horse?” The third is, “Will Intel let AMD ride its new horse?” and the fourth is, “What do we do with old horse now?”
Intel is unquestionably right on question one. The horse isn’t going to live forever, one day it’s going to have to go. AMD knows that.
They may not even disagree all that much on when the horse needs to be replaced. AMD thinks the old horse still has years of useful life ahead of it on the desktop and other less demanding tasks, but so does Intel. We wouldn’t see Itanium on the desktop until 2007 even if Intel got its way entirely.
We only get to the real problem when we get to question three: Who gets to ride the new horse. AMD is talking about how healthy the old one is, but that’s mostly because it can’t afford a new horse, and Intel won’t let them ride theirs.
If Intel offered to let AMD ride the new horse, too, in return for dropping x86-64, they’d do it. They wouldn’t look a gift horse in the mouth. 🙂 That’s quite possible the real reason why they came up with x86-64 in the first place, to eventually use it as a bargaining chip to get to ride the new nag.
Since that isn’t happening any time soon, the current dispute is really about the need to whip the old nag. AMD says, “Beat it,” while Intel up to now has said (up to now), “Leave the poor horse alone; it’s good enough.”
As you can see, it really boils down to whether you have a new horse to ride or not.
So x86-64 vs. IA-64 isn’t really a choice between A or B. It’s whether or not you whip the old horse for a while before you buy a new one. If you choose to whip the old horse, that doesn’t mean you’ve killed the new horse or the eventual need for a new horse. You’ve just delayed getting a new horse.
Even that isn’t the real reason for the dispute. That’s not the real game being played here. The deep, underlying reason for all this is who gets to ride the new horse. That’s what is driving all of this.
Intel could stop this x86-64 stuff simply by licensing IA-64 to AMD to take effect a few years down the road in return for AMD dropping it, and AMD would do it in a flash.
But I guess Intel figures it’s better (or at least cheaper) just to buy a whip and beat the horse for a while, too, especially since the new horse wasn’t meant to replace the old one for a few more years.
So Intel adopting x86-64 isn’t The Big Win for AMD, because that’s not the game. The Big Win for AMD would be Intel licensing the successor to x86 to it.
It isn’t even much of a real win, outside of practically meaningless bragging rights, because what is Intel doing? It’s stooping to conquer and taking away the only real bargaining chip and advantage AMD has over it. When the other side levels the playing field, that’s not a victory for your team.
In all likelihood, what Intel will do is come up with some sort of superset to AMD64. INT64 (to coin a term) will be able to run all AMD64 applications, but will add items that will either make INT64 apps incompatible, or more likely, somewhat less capable on AMD64 machines.
Best guess on what that would be would be something along the lines of SSE4, extra instructions not immediately supported by AMD64-based chips.
Yes, AMD will get some short-term advantages from this. Intel’s adaptation will spur x86-64 software development, and sooner rather than later.
But until the world knows how INT64 will differ from AMD64, an INT64 announcement will tend to freeze x86-64 purchases until those little details are known.
The point to all this isn’t that AMD won’t get helped by this development; it certainly will. It just won’t be helped as much, even short-term, as some might think.
In the long term, the reason why these companies exist is to generate revenues, not bragging rights. Saying, “We made Intel adopt x86-64” may turn out to be like Mac users saying “We had the first GUI.”
Or Saddam Hussein telling his American interrogators, “I made you invade Iraq.”
I’m Not Unfair, Life Is
No doubt, people are going to write me saying I’m unfair and unbiased and pro-Intel and never will say “AMD won, big time.”
Do you know what? Yes, there is unfairness and bias and pro-Intelness to what I say. But I’m not the source of all that; life is.
It is a sad fact of life that it is not fair. The strong have lots of advantages over the weak, and Intel is strong compared to AMD. I didn’t make Intel strong or AMD weak; I didn’t give strongness all the advantages it has against weakness.
All I do is notice that the playing field is uneven, and point out the effects of that.
One of the advantages of being strong is that you can turn lemons into lemonade a lot easier than if you’re weak. Intel can and does make mistakes that would kill AMD, but they’re strong enough to shrug them off.
Being strong makes it much easier to use business jujitsu and clobber the weak with their own moves. x86-64 is a good move on AMD’s part, but Intel can neutralize it simply by adapting it, and in all likelihood, the market isn’t going to give AMD much of a reward for getting their first.
Sure, you can make an argument that it is unfair, but it doesn’t matter in the real world. What is matters, not what is fair.